Four of the 22 Easy Chat groups carry no text of their own: POKéMON,
POKéMON (NAT), MOVE 1 and MOVE 2 are lists of species and move ids, and
pret prints them through gSpeciesNames and gMoveNames. The extractor
resolves them the same way, so easy_chat_data.lua ended up holding 807
copies of names the dataset already has -- copies that go stale the
moment anything renames a species or a move. A mod that renames them,
a translation filling the species_names and move_names catalogs first
of all, changed every screen but this one: the picker still offered
the cart's English names, and the profile read back in them.
Resolve those two kinds of word at draw time instead, against the same
Pokemon.name / Pokemon.moveName the party screen and the Pokédex use.
When the dataset has no entry for an id it answers with a placeholder
("POKéMON 063", "-------"), which is worse than the word the cart
shipped, so that case keeps the extracted text. An entry that names
the group ("easyChat.POKéMON|PIKACHU") still wins over both, since a
species name may need a wording other screens do not use. A bare-key
entry does not: bare keys are shared with menu labels this engine
already translates (CUT, FLASH, STRENGTH), and one of those should not
decide what a move is called here.
tools/modkit.py keeps emitting those 807 rows, since the species and
move catalogs it writes come from the Gen 1 dataset and do not list a
Hoenn species or a Gen 3 move; its note now says an empty entry shows
the game's own name, so a translator only fills the ones nothing else
renames.
The Easy Chat screen picks its words from src/core/game3/easy_chat_data.lua, the ROM's own word tables as the extractor writes them: 1,813 English words across 22 groups, plus the group names the picker lists. It drew them straight from that table, so every word the player composes with stayed English, and the screen's own prompts -- routed through Strings() already -- sat in a frame of untranslated vocabulary.
src/core/game3/easy_chat_text.lua is the seam. The data module stays exactly what the extractor produced, because it is regenerated from the ROM and anything written into it would be lost on the next import; the words are translated where they are drawn instead, the same split the rest of game3 uses for extracted text. Five sites go through it: the group list, its header, the word list, the word sitting in a slot, and the saved profile a message box prints (ShowEasyChatMessage). The ids the save file stores are untouched.
Each word carries its group's context ("easyChat.FEELINGS"), because the same English word means different things in different groups and 155 of them collide with labels this engine already translates elsewhere (ATTACK, BAG, CANCEL, BACK...). Strings() falls back to the plain key, so a catalog that does not need the distinction lands with one entry -- with the consequence that a mod which already translates BAG for the menus now sees that wording in the Easy Chat list too, until it adds a context-specific entry.
A seam nothing can discover is not much use: tools/modkit.py's catalog generator emits these keys too. The literal harvester cannot see them (the picker looks each word up at draw time, not at a Strings("...") call site), so the dataset dump walks EasyChatData.GROUPS and writes 1,835 rows into lang/strings.lua, where a translator finds them beside the rest of the engine text.
Each word is clipped to the box it sits in: in the picker, the red selection rectangle; in the phrase frame, the slot's own frame. The cart's words fit those boxes (72 px at most); a longer translation is cut at the box edge instead of running past the highlight, into the next slot's cursor, or off screen. Group names get the room up to the scroll arrows instead, because an official translation already needs more than the rectangle: the French cart's VIE QUOTIDIEN. is 83 px.
tests/engine/game3_easy_chat_text_test.lua covers the context lookup, the fallback to a plain key, the untouched English with no catalog, and that the extracted data itself is never rewritten. It lives in tests/engine so the T1/T2 tier actually runs it.
There was no way to ask "will this mod load on FireRed?". `validate`, `lint`
and `pack` only reason about Gen 1, and `gen2check` refuses a Gen 3 mod at the
manifest gate. `gen3check` is the Gen 3 counterpart of `gen2check`:
python3 tools/modkit.py gen3check mods/<id> [--notes] [--json] [--strict]
- Factor the shared compat analysis behind a `Generation` descriptor, with
`GEN2`/`GEN3` module-level singletons holding everything that differs: the
compat facade name, the generation's own directory (`gen2` vs `game3`), the
legacy manifest flag, the screen-twin prefix, and the sibling spelling. Every
compat helper now takes a trailing `gen`.
- MK400-MK410 stay shared by design -- they are the same questions asked of a
different facade -- and the verdict line (`on gen 3: ...`) disambiguates.
- Gen 3 sibling spelling: `src/ui/game3/` is snake_case while
`src/world/game3/WorldAPI.lua` is not, so MK403 tries both spellings and
stops at the first sibling that exists. Only 8 Gen 1 modules are reachable
this way (`BattleAPI`, `BagMenu`, `HallOfFame`, `IntroMovie`, `OptionRows`,
`ShopMenu`, `SummaryMenu`, `TrainerCard`), so MK403 on Gen 3 is live but
narrow.
- MK409's screen-twin half is switched off on Gen 3 -- there is no
`Screens.GEN2_IDS` equivalent -- while its version-string half still runs,
since a Gen 1 version id in a mod that declares only FireRed is a real bug.
- Rename `GEN2_IDS_DUMP` to `VERSION_IDS_DUMP`, now parameterized by
generation.
Gen 2 behaviour is unchanged: `diff -r` of `gen2check --notes` and
`gen2check --notes --json` output over all 10 shipped mods, captured before and
after this change, is empty.
Tests: `tests/modkit/cases/gen3check.lua` grows from 504 to 542 checks. Every
fixture is derived from the engine at run time -- the coverage table,
`GEN1_ONLY_MODULES`, and the `game3/` directory listings -- so the suite cannot
drift away from the tables it is asserting against. It covers MK400's Gen 3
wording, MK402 naming `Gen3Compat` and never `Gen2Compat`, MK403 naming the
snake_case path, MK404 quoting the member, MK409's version-string half, MK410
counted once at file scope, the `--json` verdict shape, and that no adapted
module is ever MK402.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Crystal boots from a user-supplied ROM, imports a full cache and is playable:
copyright, the Crystal intro movie, the animated title, gender select, Oak,
and out into Johto. 122 of the cart's 169 script specials are implemented.
Import and data
- tools/make_crystal_manifest.py derives the manifest by importing
make_gold_manifest as a library, with three additive keyword seams. Gold and
Silver still regenerate byte-identical, which is the standing requirement for
touching that generator.
- crystal_symbol_deltas.py and crystal_movie_symbols.py carry the symbol delta:
Crystal renames the credits mons, splits the trainer card, Pokegear and
pack-pal blocks by gender, and replaces the intro and title outright.
- Crystal-only manifest keys: engineFlagOrder (162 flags to Gold's 93, so the
badge block sits one higher) and unownCharmap (the main charmap parser stops
at the first newcharmap so the two cannot contaminate each other).
Extractor
- RomExtractorGen2 becomes three-edition. Crystal corrections: PAL_MAP_BANK
0x13, a flat PICS_FIX pic bank, audio bank 0x5e, the mapSongs id-100 hole,
seven NPC trades, a TradeTexts stride of 8, the five Crystal tileset anim
steps with per-row degrade, and the column-major trainer card portraits.
- New: animated front sprites (frames, bitmasks, play and idle scripts), the
Battle Tower roster, Kris assets, Mobile System GB art, and the Crystal
intro and title via src/import/CrystalMovie.lua.
Engine
- GameVersion gains engine(id) and fixes(id). Gold and Silver keep their
original bugs where the bug is not hardware dependent; Crystal gets the fixes
Crystal shipped: Lucky Number boxes 10-14, surfing onto an NPC, and the
Reflect and Light Screen defence overflow.
- Crystal story: Suicune and Eusine, Celebi behind the GS Ball flag, the Ruins
of Alph chambers, Buena, the Move Tutor, the Poke Seer, and the Battle Tower
including the wInBattleTowerBattle badge-boost guard.
- Kris and the gender flag, animated fronts in battle and the summary screen,
and mon caught data.
Verification
- Every extracted asset is pixel-compared against pret's own source PNGs.
- Gold caches are byte-identical before and after, file for file.
- New Crystal suites plus a T2 Gen 2 tier; the full suite passes.
Silver: derived import manifest (tools/make_silver_manifest.py re-resolves
the Gold manifest's symbols from pokesilver.sym), silver GameVersion row,
generation-keyed extractor routing, required-files override, edition save
stamping (a Silver playthrough no longer writes into the Gold save),
checkver-driven edition data, SILVER/KAMON/OSCAR/MAX presets, GOLD rival
default, edition credits banner, Lugia title screen (OAM layouts, bob,
trail, palettes as title.lua data keys with Gold defaults so old caches
need no re-import), packaging for every build target, docs, and tests.
Launcher: the installed-mods list is one continuous scroll (rows culled to
the viewport) instead of a pager with an inner scroll viewport; the pad
cursor's edge-scroll no longer runs it to the bottom. The game dropdown
shows just the initial and caret. Find-tab behavior unchanged.
Title tempo: a sprite-anim frame shows duration+1 ticks
(engine/sprite_anims/core.asm GetSpriteAnimFrame), which locks both
editions' 64-tick wing beat to the 64-tick sine bob; the title screens no
longer run fast and out of phase.
A fixed cutoff (e.g. "<=200 is ink") only makes sense for sprites with a light background to split against; a mostly-opaque 16x16 icon has almost no pixel above that cutoff, so every such icon collapsed onto the same "all ink" hash and was flagged as a near-duplicate of anything else that also collapsed -- which was most of them, boulder.png included. Thresholding against the image's own mean keeps the split meaningful (and roughly balanced) no matter how light or dark the source is.
subprocess.run(..., capture_output=True, text=True) with no explicit
encoding falls back to locale.getpreferredencoding(False) -- the OS
default codepage. On Windows that's a legacy single-byte codepage
(e.g. cp1252), never UTF-8. When the LuaJIT dump contains a byte with
no mapping in that codepage, subprocess's internal _readerthread
crashes with an uncaught UnicodeDecodeError in a background thread; the
thread dies silently and the caller gets back stdout=None instead of a
string, crashing one line later with
AttributeError: 'NoneType' object has no attribute 'splitlines'.
Concretely, the Yellow-side imported dataset contains:
"_ColosseumHeightText" -> "...6’8” tall!"
The right double quotation mark (U+201D) encodes in UTF-8 as E2 80 9D;
0x9D has no defined character in cp1252, so decoding as cp1252 fails
outright. Verified against the real imported dataset: the Red/Blue-only
dump has zero bytes outside cp1252's defined range; the Yellow dump has
exactly one, at this row.
UTF-8 is the actual encoding these dumps are produced in -- the driver
Lua sources are read/written as UTF-8 throughout this file, and LuaJIT
writes those source strings' bytes back out verbatim -- so passing
encoding="utf-8" explicitly at the three affected call sites
(run_loader, check_data_dump, dump_dataset) is a no-op on platforms
whose default codepage is already UTF-8 (Linux/macOS) and a correctness
fix on Windows.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This fixes the nil value returned when running validate or pack on Fedora 43. Since Love isn't running, luajit calls on an empty table. Providing a stub table resolves the nil error.
bryanthaboi